Instructions for Term Paper: “Mini Meta-Analysis”

Select an epidemiological association of interest to you, regarding a specific health condition and a factor that either protects or increases the risk for that condition (excluding any examples used in the meta analysis lecture). Using an example from that lecture, a valid research question might be: “Does fruit and vegetable consumption decrease the risk for breast cancer?” You are required to search for relevant articles on the PubMed website: remember to access it via the Dahlgren Library or other Georgetown library web site, so that you can get the full text articles, which you will need in order to do this assignment. The rules for article selection are:

  1. Full text, original research articles only – no reviews, and no meta-analyses
  2. English language
  3. Case-control, cohort, or randomized controlled studies only
  4. Select a minimum of 8 articles and a maximum of 12
  5. See instructions in Results C.3 below, i.e. be sure that your selected set of articles will allow you to conduct one subset analysis by study design, year, or other characteristic such as gender.

Preliminary search and approval process

Don't use plagiarized sources. Get Your Custom Essay on
Instructions for Term Paper: “Mini Meta-Analysis”
Just from $13/Page
Order Essay

Conduct a preliminary search to make sure enough articles that meet the above criteria have been published for your research question. If there are more than 12 articles published, come up with at least two different ways of how you will choose the final 12 (most recent, largest sample sizes, study design type, etc. – see Methods section on Page 2 of this document). Prepare a document with the following information and bring a print copy to class on April 4:

  • Study question
  • PubMed search terms used
  • Number of relevant articles found based on preliminary search
  • A plan for choosing the final set of articles (if >12 relevant articles have been published)

You can continue with your term paper once you receive final approval from Dr. Dash or Dr. Loffredo, both of whom will be in the classroom to discuss your term papers. DO NOT START WRITING YOUR FINAL TERM PAPER OR MAKING TABLES BEFORE RECEIVING WRITTEN APPROVAL FROM A FACULTY MEMBER IN THIS COURSE. If changes are recommended in your topic, please follow up with them by email.

Once you have received approval, read each article carefully and thoroughly to ensure that it meets the above criteria. Sometimes a paper appears to be relevant according to the abstract, but when you read the full text you find problems. If that happens, you may need to exclude the paper and go on to the others. Save the PubMed abstract to a file, which you will include in your paper (see below). Be sure that you understand the methods and results, and that you can extract the OR or RR and the 95% CI from the article. Sometimes you will find an article that does not explicitly report those statistics but does report the data such that you can calculate it yourself: this would be a valid study to include if all other criteria are fulfilled. You will be extracting key information from each article and entering into a table, as described below; you will also create forest plots and write a qualitative assessment of the strengths and weaknesses of the articles you included. Therefore this assignment combines elements of the narrative (qualitative) review and the meta-analysis (quantitative).

Required Elements of the Paper

In general, the paper must be double-spaced, 11- or 12-point font, with all pages numbered, and with your name included in the header on each page. It should be a single document with embedded tables and graphs. There are no specifications for margins. Please observe the page limits described below.

Title page: one page

Write the title of your paper (e.g. “Association of marigolds with increased risk for chicken allergy”). Include the following statement and sign it.

I pledge on my honor that I have not given nor received any unauthorized assistance on this term paper, nor have I read or included any previously published meta-analysis on my research topic.

___________________                  _______________________________        __________

Name                                                  Signature                                             Date

Introduction: one page

Give a general overview of the health condition and the risk/protective factor that you are investigating. State the specific research question you are investigating. If the results of your analysis show that the factor appears to be strongly associated with the disease, how might the public health be impacted?

Methods: one page

Describe the specific search terms that you used to find your articles on PubMed. How many articles did that search yield, and how did you narrow your selection to 8 to 12 articles? It is very important to avoid bias in how you selected the articles to review. If you are not sure how to limit your article set, ask any of the faculty for guidance. (For example, you found 20 articles using the search terms “marigolds AND chicken allergy” and selected the 10 most recent papers. Or you found 30 and selected only the 12 most recent studies that were conducted in the U.S.) Be aware that the faculty should be able to repeat your search and find the same initial number of articles, and the subset that corresponds to your selection criteria.

Describe the specific methods that you used to compare and contrast results between studies, e.g. what graphics and sub-group analyses did you conduct? Mention the specific software you used to create the forest plots.

Results: no page limits, suggested maximum of 8 (4 for tables and graphs, 4 for text)

Create a table listing the articles you analyzed. Each row will be one specific article. Make columns for the first author’s last name, year of publication, location (country, state, etc…), type of study design (case-control, etc…), sample size, method of assessment of the health condition, method of assessment of the risk or protective factor, the unadjusted OR or RR and its 95% CI, and the adjusted statistic and what was included in the adjustment (for example, RR=2.0, 95% CI 1.5-3.5, adjusted for age and gender). You can include additional columns for any other factors or subgroups you think are important (optional). It is not necessary to order the studies by year unless that makes sense for your specific topic; you can place them into any order that helps you to show important findings, e.g. by study design. You must write descriptive text (one or more paragraphs) to summarize the major results shown in your table: not describing every row and column, but giving your overall impressions. Embed the table following the descriptive text.

Create a forest plot for the full group of studies you analyzed in the table (a list of available software will be provided separately, or you can make the plot by hand or in Excel). Clearly label all elements of the plot, as in the example from the lecture. Write a paragraph to summarize your impressions of the pattern revealed in the plot. You do not have to calculate the summary statistic and 95% CI, nor the Q statistic. It is not necessary to make each data point proportionate to its sample size. Embed the plot following this descriptive text.

Create a minimum of two additional forest plots for a subset analysis. For example, compare and contrast findings by case-control vs. cohort studies, by year of publication, by method of disease or risk factor assessment, or by any other characteristic that you believe to be important (e.g. gender, race, age group). Write one or more paragraphs describing your impressions/interpretations of the data plots. Embed the plots following this descriptive text.

Summary and Conclusions: one or two pages

Briefly summarize what was revealed in your analysis. Then comment on the relative strengths and weaknesses of the studies, not article by article, but making more general observations across studies (e.g. “Six of the ten studies used a very flawed questionnaire to estimate marigold consumption”). Be sure to use the “epidemiological” terminology and concepts we have learned in the course (show your knowledge!), e.g. the types of bias and confounding that might exist given the study design, and whether these appeared to be actually present in the specific studies you analyzed. Speculate on the future direction(s) that research should ideally take to more conclusively answer the general research question you posed in the Introduction. Conversely, are you satisfied with the results of your analysis, and why? Finally, provide a one-sentence Conclusion that answers your research question.

Appendix: no page limits

Attach the downloaded PubMed abstracts of your articles. Hint: you can save abstracts to the clipboard in PubMed and then you can save to file or e-mail to yourself. (Note: if you need help with PubMed techniques in general, campus library staff members are very skilled and can give you some pointers.)


and taste our undisputed quality.